Adding AI false alarm filtering to Milestone XProtect

Article, Partner

October 22, 2026

DeepAlert

A mature Milestone installation usually has a lot invested in it already.

Cameras have been configured. Events and rules have been built. Operators know the interface. Processes have been developed around how alarms are received and handled.

So when an ARC, integrator or end user wants to add AI video analytics, replacing any of that makes little sense.

The more practical question is: can you improve the quality of the events reaching operators without changing the VMS they already use?

That is where DeepAlert Core fits into a Milestone environment. It filters 95% of nuisance alerts to dramatically reduce operator workload and improve accuracy and efficiency of monitoring operations.

Keep Milestone at the centre

DeepAlert Core does not replace Milestone. It adds a cloud-based AI filtering layer to the existing deployment.

A lightweight Windows connector is installed on the Milestone server or on a machine alongside it. Milestone sends events across the local network to the connector, which securely sends the relevant event images to the DeepAlert cloud over HTTPS for analysis.

DeepAlert analyses those images to identify whether a person or vehicle is present, then returns the result to the Milestone dashboard.

Milestone remains the operator’s working environment.

Cloud AI architecture

The architecture also matters from an IT and deployment perspective.

DeepAlert does not require the camera estate to be replaced or upgraded to specific AI-enabled cameras.

Raw video remains on site. The connector sends only the event snapshots and clips required for analysis to the DeepAlert cloud, rather than continuously sending the site’s video streams.

Communication from the connector uses outbound HTTPS.

The DeepAlert connector currently supports XProtect 2026 R2.

Why this matters on larger camera estates

The value becomes clearer as the number of cameras and events increases.

A camera estate can grow faster than the team responsible for monitoring it. More cameras mean more events, but not necessarily more useful information.

Core adds a filtering step between the event being generated and the operator reviewing it. It analyses whether the event contains the person or vehicle the operator actually needs to investigate.

Add intelligence without rebuilding the stack

For established Milestone users, adding AI should not mean starting again.

For ARCs, integrators and security teams already invested in Milestone XProtect, that is the real value of the integration.

To see DeepAlert Core working with Milestone XProtect, book a demo at deepalert.ai or contact [email protected].

Article

Day in the Life: Bohdan Tokarskyi, Frontend Lead Developer

Bohdan Tokarskyi came to software from construction engineering. Today he leads DeepAlert’s frontend team, splitting his week between writing code and reviewing it, and working on DeepAlert COMMAND, the platform he built from scratch. He talks about what changed when he stepped into leadership, what good design has to do six hours into an operator’s shift, and how he spends his time away from the screen.

Article

DeepAlert Welcomes Manguard Plus

DeepAlert welcomes Manguard Plus as a new client. Our AI video analytics will support their 24/7 Command & Control Centre by filtering unnecessary CCTV activations, helping operators focus on genuine security events and supporting a more effective hybrid security approach.

Article

What does improving an AI video analytics model actually look like?

An AI video analytics model must adapt to changing cameras, environments and conditions. We look at how DeepAlert investigates challenging scenarios, refines its models and tests improvements to deliver more consistent, reliable detection in real-world surveillance.